Cost-effective Instruction Learning for Pathology Vision and Language Analysis

TL;DR

CLOVER is a cost-effective, lightweight framework for pathology vision-language analysis that leverages instruction tuning with GPT-3.5 prompts, outperforming larger models in clinical question-answering tasks.

Contribution

The paper introduces CLOVER, a novel lightweight instruction learning framework that reduces training costs by freezing large model parameters and using GPT-3.5 prompts for pathology analysis.

Findings

CLOVER outperforms larger models with 37 times more parameters.

Hybrid instructions improve visual question-answering in pathology.

CLOVER demonstrates robustness in few-shot clinical dataset adaptation.

Abstract

The advent of vision-language models fosters the interactive conversations between AI-enabled models and humans. Yet applying these models into clinics must deal with daunting challenges around large-scale training data, financial, and computational resources. Here we propose a cost-effective instruction learning framework for conversational pathology named as CLOVER. CLOVER only trains a lightweight module and uses instruction tuning while freezing the parameters of the large language model. Instead of using costly GPT-4, we propose well-designed prompts on GPT-3.5 for building generation-based instructions, emphasizing the utility of pathological knowledge derived from the Internet source. To augment the use of instructions, we construct a high-quality set of template-based instructions in the context of digital pathology. From two benchmark datasets, our findings reveal the strength of hybrid-form instructions in the visual question-answer in pathology. Extensive results show the cost-effectiveness of CLOVER in answering both open-ended and closed-ended questions, where CLOVER outperforms strong baselines that possess 37 times more training parameters and use instruction data generated from GPT-4. Through the instruction tuning, CLOVER exhibits robustness of few-shot learning in the external clinical dataset. These findings demonstrate that cost-effective modeling of CLOVER could accelerate the adoption of rapid conversational applications in the landscape of digital pathology.